Device, Method, and Medium for Establishing Language Model for Expanding Finite State Grammar Using a General Grammar Database

1. A device for establishing a language model, comprising: a schema-expander to expand a state schema which is composed of one or more states defined by a finite state grammar using a general grammar database; a grammatical-structure-expander to expand grammatical structures which can be expressed by each state of the expanded state schema using the general grammar database; a grammatical-structure-filter to filter out any incorrect grammatical structure from the expanded grammatical structures using the general grammar database, wherein the grammatical-structure-filter includes a property value comparing module to compare property values of a first conjunction node and a second conjunction node that compose a conjunction node of a conjunction state generated through substitution and adjunction; and a filtering module to assign only an intersection of the property values of the first and second conjunction nodes as a property value of the conjunction state, and to store the conjunction state in the general grammar database if the first conjunction node or the second conjunction node has two or more property values.

2. The device of claim 1 , wherein the head state is a predicate.

3. The device of claim 1 , wherein the general grammar database has tree structure data that includes data concerning structures and meanings of phrases or clauses, words included in the structures, and parts of speech corresponding to the words.

4. The device of claim 3 , wherein the schema-expander includes: a state classifying module to classify each state into the head state having a main meaning in a sentence or the dependent state which is any state excluding the head state; a grammar searching module to search the general grammar database to detect tree structure data having a state identical to the classified head state in a root node; and a comparing/expanding module to compare the number of sentence components in the detected tree structure data with the number of dependent states added to the head state and to expand the state schema according to the comparison results.

5. The device of claim 4 , wherein the state classifying module includes: a tag-information extractor to extract tag information including state information from the states; a phrase-information extractor to extract phrases that can be expressed by the states only or by the connection between the states from the extracted tag information; and a classifier to classify each state into the head or dependent state using the phrase information extracted by the phrase information extractor.

6. The device of claim 4 , wherein the comparing/expanding module: inserts states corresponding to the sentence components in the detected tree structure data into the dependent states if the number of the sentence components is greater than that of the dependent states; and deletes states unused in the tree structure data from the dependent states if the number of sentence components is smaller than that of the dependent states.

7. The device of claim 6 , wherein the grammatical-structure-expander includes: a head-state-expanding module to expand a grammatical structure that can be expressed by a head state included in the expanded state schema using the general grammar database; and a dependent state expanding module to expand a grammatical structure that can be expressed by a dependent state modifying the expanded head state using the general grammar database.

8. The device of claim 7 , wherein the expansion includes a substitution process for substituting the head or dependent state by tree structure data corresponding to the same part of speech and an adjunction process for adding the tree structure data between the states.

9. The device of claim 8 , wherein the grammatical-structure-filter further includes: a filtering module to determine whether the two property values of the first and second conjunction nodes are identical, to assign the identical value as a property value of the conjunction state and store the conjunction state in the general grammar database if the two property values are identical, and to recognize that there was a failure in the substitution and adjunction processes and not store the conjunction state in the general grammar database if the property value of the first conjunction node is different from that of the second conjunction node.

10. A method of establishing a language model, comprising: (a) expanding a state schema which is composed of at least one state defined by a finite state grammar using a general grammar database; (b) expanding grammatical structures which can be expressed by each state of the expanded state schema using the general grammar database; and (c) filtering out any incorrect grammatical structure from the expanded grammatical structures using the general grammar database, wherein (c) includes (c1) comparing property values of a first conjunction node and a second conjunction node that compose a conjunction node of a conjunction state generated through substitution and adjunction and (c2) assigning only an intersection of the property values of the first and second conjunction nodes as a property value of the conjunction state, and storing the conjunction state in the general grammar database if the first conjunction node or the second conjunction node has two or more property values.

11. The method of claim 10 , wherein the general grammar database has tree structure data that includes data concerning structures and meanings of phrases or clauses, words included in the structures, and parts of speech corresponding to the words.

12. The method of claim 11 , wherein (a) includes: (a1) classifying each state into the head state having a main meaning in a sentence or the dependent state which is any state excluding the head state; (a2) searching the general grammar database to detect tree structure data having a state identical to the classified head state in a root node; and (a3) comparing the number of sentence components in the detected tree structure data with the number of dependent states added to the head state and expanding the state schema according to the comparison results.

13. The method of claim 12 , wherein (a1) includes: extracting tag information including state information from the states; extracting phrases that can be expressed by the states only or by the connection between the states from the extracted tag information; and classifying each state into the head or dependent state using the phrase information extracted by the phrase information extractor.

14. The method of claim 12 , wherein (a3) includes: inserting states corresponding to the sentence components in the detected tree structure data into the dependent states if the number of the sentence components is greater than that of the dependent states; and deleting states unused in the tree structure data from the dependent states if the number of sentence components is smaller than that of the dependent states.

15. The method of claim 14 , wherein (b) includes: (b1) expanding a grammatical structure that can be expressed by a head state included in the expanded state schema using the general grammar database; and (b2) expanding a grammatical structure that can be expressed by a dependent state modifying the expanded head state using the general grammar database.

16. The method of claim 15 , wherein the expansion includes a substitution process for substituting the head or dependent state by tree structure data corresponding to the same part of speech and an adjunction process for adding the tree structure data between the states.

17. The method of claim 16 , wherein (c) further includes: (c2) determining whether the two property values of the first and second conjunction nodes are identical, assigning the identical value as a property value of the conjunction state and storing the conjunction state in the general grammar database if the two property values are identical, and recognizing that there was a failure in the substitution and adjunction processes and not storing the conjunction state in the general grammar database if the property value of the first conjunction node is different from that of the second conjunction node.

18. At least one computer readable medium storing instructions that control at least one processor to perform a method comprising: (a) expanding a state schema which is composed of at least one state defined by a finite state grammar using a general grammar database; (b) expanding grammatical structures which can be expressed by each state of the expanded state schema using the general grammar database; and (c) filtering out any incorrect grammatical structure from the expanded grammatical structures using the general grammar database, wherein (c) includes (c1) comparing property values of a first conjunction node and a second conjunction node that compose a conjunction node of a conjunction state generated through the processes of substitution and adjunction and (c2) assigning only an intersection of the property values of the first and second conjunction nodes as a property value of the conjunction state, and storing the conjunction state in the general grammar database if the first conjunction node or the second conjunction node has two or more property values.

19. At least one computer readable medium as recited in claim 18 , wherein the general grammar database has tree structure data that includes data concerning structures and meanings of phrases or clauses, words included in the structures, and parts of speech corresponding to the words.

20. At least one computer readable medium as recited in claim 19 , wherein (a) includes: (a1) classifying each state into the head state having a main meaning in a sentence or the dependent state which is any state excluding the head state; (a2) searching the general grammar database to detect tree structure data having a state identical to the classified head state in a root node; and (a3) comparing the number of sentence components in the detected tree structure data with the number of dependent states added to the head state and expanding the state schema according to the comparison results.

21. At least one computer readable medium as recited in claim 20 , wherein (a1) includes: extracting tag information including state information from the states; extracting phrases that can be expressed by the states only or by the connection between the states from the extracted tag information; and classifying each state into the head or dependent state using the phrase information extracted by the phrase information extractor.

22. At least one computer readable medium as recited in claim 20 , wherein (a3) includes: inserting states corresponding to the sentence components in the detected tree structure data into the dependent states if the number of the sentence components is greater than that of the dependent states; and deleting states unused in the tree structure data from the dependent states if the number of sentence components is smaller than that of the dependent states.

23. At least one computer readable medium as recited in claim 22 , wherein (b) includes: (b1) expanding a grammatical structure that can be expressed by a head state included in the expanded state schema using the general grammar database; and (b2) expanding a grammatical structure that can be expressed by a dependent state modifying the expanded head state using the general grammar database.

24. At least one computer readable medium as recited in claim 23 , wherein the expansion includes a substitution process for substituting the head or dependent state by tree structure data corresponding to the same part of speech and an adjunction process for adding the tree structure data between the states.

25. At least one computer readable medium as recited in claim 24 , wherein (c) further includes: (c2) determining whether the two property values of the first and second conjunction nodes are identical, assigning the identical value as a property value of the conjunction state and storing the conjunction state in the general grammar database if the two property values are identical, and recognizing that there was a failure in the substitution and adjunction processes and not storing the conjunction state in the general grammar database if the property value of the first conjunction node is different from that of the second conjunction node.